Find results with:

Add another field

Search by date

to

Searching collections:

Dartmouth Dissertations

Add or remove collections

Home Dartmouth Dissertations mao2012 1

Reference URL

To link to the entire object, paste this link in email, IM or document

To embed the entire object, paste this HTML in website

To link to this page, paste this link in email, IM or document

To embed this page, paste this HTML in website

State representation for agent-based reinforcement learning.

you wish to report:

Your comment:

Your Name:

Submit

Cancel

...

Back to top

Powered by CONTENTdm®

Author	Mao, Tao
Title	State representation for agent-based reinforcement learning.
Date	2012
Abstract	Agent-based policy learning in complex and uncertain environments is challenged by escalating computational complexity with the size of the task space (action choices and environmental states) as well as the number of agents. Nonetheless, there is ample evidence in the natural world that high functioning social mammals learn to solve complex problems with ease, both individually and cooperatively. Neuroscientific research shows that brain structures embed computational functions that include state abstraction, hiearchical state representation, and action space chaining. Drawing from these themes, this thesis develops two important concepts for reducing the state complexity: state abstraction (mapping an external state vector to an internal state representation of lower dimension) and hierarchical state representation (representing a state space by a hiearachy of subspaces thereby reducing the number of states used to represent the environment). The Q-Tree algorithm, which provides simultaneous state abstraction, hierarchial representation, and policy learning in a single algorithm, is introduced, requiring no prior knowledge of candidate boundaries for separating the state space into subspaces when solving agent-based reinforcement learning problems; The Q-Tree algotithm automatically constructs linear separations on multiple dimensions and has a computational complexity linear in the dimension of the state vector and independent of the potential number of state boundaries, which is superior to existing methods of constructing hierarchical state representations for reinforcement learning. These methodologies are applied to single- and multi-agent task domains such as foraging, patrolling, herding, and path planning so as to demonstrate preformance in terms of convergence rate, computational efficiency, memory requirements and generalizability of learned policies. The Q-Tree algorithm is also employed to model incubation and restructuring in insight problem solving, accounting for three-stage learning curves observed in rhesus monkeys in a reverse-reward contingency task and generalizing learned knowledge to new but similiar situations.
Description	xv, 185 leaves : illustrations (some color)
Format	application/pdf
Advisor	Ray, Laura E. ; Amato, Christopher ; Phan, Minh Q. ; Kralik, Jerald D. ; Granger, Richard
College	Thayer School of Engineering.
Issuing Body	Dartmouth Dissertations are made available through the Dartmouth Digital Library.
Degree	Thesis (Ph.D)--Dartmouth College.
Copyright	In Copyright. This material is protected by copyright and/or related rights. You are free to use this material in any way that is permitted by copyright law that applies to your use. For other uses, you need to obtain permission from the rights-holder(s). http://rightsstatements.org/page/InC/1.0/
DOI	https://doi.org/10.1349/ddlp.3529

Title	mao2012 1
Full Text	STATE REPRESENTATION FOR AGENT-BASED REINFORCEMENT LEARNING A Thesis Submitted to the Faculty in partial fulfillment of the requirements for the degree of Doctor of Philosophy By Tao Mao Thayer School of Engineering Dartmouth College Hanover, New Hampshire September 2012

Find results with:

Searching collections:

Home Dartmouth Dissertations mao2012 1

State representation for agent-based reinforcement learning.

Object Description

Description

State representation for agent-based reinforcement...

State representation for agent-based reinforcement...

	All
	Pages Enter page numbers or page ranges separated by commas. For example: 1,3,5-12